Mario Ciampelli, Fulvio Leoni, Francesco Cucinelli, Salvatore Mancuso, Simona Panunzi, Andrea De Gaetano, and Antonio Lanzone

Transcription

1 X/05/$15.00/0 The Journal of Clinical Endocrinology & Metabolism 90(3): Printed in U.S.A. Copyright 2005 by The Endocrine Society doi: /jc Assessment of Insulin Sensitivity from Measurements in the Fasting State and during an Oral Glucose Tolerance Test in Polycystic Ovary Syndrome and Menopausal Patients Mario Ciampelli, Fulvio Leoni, Francesco Cucinelli, Salvatore Mancuso, Simona Panunzi, Andrea De Gaetano, and Antonio Lanzone Department of Obstetrics and Gynecology, Università Cattolica del Sacro Cuore (M.C., F.L., F.C., S.M.), Rome, Italy; Laboratorio di Biomatematica, Consiglio Nazionale delle Ricerche Istituto di Analisi dei Sistemi ed Informatica (S.P., A.D.G.), Rome, Italy; and OASI Institute for Research (A.L.), Troina, Italy INSULIN IS THE most potent endogenous anabolic hormone and is responsible for storage of metabolic fuels. Among its numerous actions, it is characteristically recognized for its ability to stimulate glucose uptake into insulinsensitive tissues (1). Insulin resistance, particularly when associated with central obesity, is now recognized as part of a syndrome associated with a number of cardiovascular risk factors, such as dyslipidemia, hypertension, dysfibrinolysis, and glucose intolerance (2). Insulin resistance syndrome, originally described by Reaven (2), is associated with a greater than 2-fold relative risk for myocardial infarction and macrovascular disease (3). First Published Online December 14, 2004 Abbreviations: AUC, Area under the curve; BMI, body mass index; CVD, cardiovascular disease; HOMA, homeostasis model assessment; IGT, impaired glucose tolerance; NPV, negative predictive value; OGTT, oral glucose tolerance test; PCOS, polycystic ovary syndrome; PPV, positive predictive value; ROC, receiver operating characteristic. Sib, Avignon insulin sensitivity basal index; SiM, Avignon insulin sensitivity index. JCEM is published monthly by The Endocrine Society ( endo-society.org), the foremost professional society serving the endocrine community Background: Polycystic ovary syndrome (PCOS) and menopausal subjects are characterized by an increased cardiovascular and type 2 diabetes mellitus risk, at least partially related to insulin disturbances. The evaluation of insulin resistance in these patients could be useful as primary prevention. The aim of the study was to verify the validity of several indexes of insulin sensitivity in PCOS and menopausal subjects by comparing the data obtained by these indexes to those of euglycemic-hyperinsulinemic clamp studies. Methods: One hundred PCOS and 110 menopausal subjects were analyzed; all subjects underwent an oral glucose tolerance test (75 g) and euglycemic-hyperinsulinemic clamp study. Seven PCOS patients and 13 menopausal subjects had impaired glucose tolerance or type 2 diabetes mellitus and were excluded from the study. After analysis of correlation coefficients between the evaluated indexes and the clamp studies, the sensitivity and specificity of different cut-off values for each parameter were analyzed by receiver operating characteristic (ROC) curves. Results: The best correlation coefficients with clamp studies were obtained with the Avignon insulin sensitivity index (SiM) (R s ) in PCOS patients and the Matsuda and De Fronzo index (R s ) in menopausal patients. The best predictive index of insulin resistance in PCOS was a Avignon insulin sensitivity basal index (SibB) value of 62 or less (78% sensitivity, 95% specificity) and an insulin area under the curve (AUC) of 7,000 IU/ml or more (>50,225 pmol/ liter) 120 min (83% sensitivity, 90% specificity). In the menopausal population, the best predictive performance was obtained by an insulin AUC of 10,000 IU/ml or more (>71,750 pmol/liter) 240 min (70% sensitivity, 88% specificity). Conclusions: The presence of high correlation coefficients does not necessarily mean that the indexes of insulin resistance have an optimal predictive performance; this is probably due to the presence of many borderline values. The simple evaluation of insulin AUC seems to effectively replace the euglycemic-hyperinsulinemic clamp in routine clinical practice, allowing results superimposable to those obtained by minimal model analysis. (J Clin Endocrinol Metab 90: , 2005) In women s care clinical practice there are two kinds of patients who should be classified as moderate-high cardiovascular disease (CVD) risk subjects, who are often characterized by impaired glucose tolerance (IGT) or frank diabetes mellitus (type 2): menopausal and polycystic ovary syndrome (PCOS) patients. In both of these groups several alterations of insulin metabolism might play a key role in increasing the risk for CVD and type 2 diabetes mellitus. Menopause is characterized by a progressive detrimental effect on insulin-mediated glucose disposal, as suggested by the presence of a negative correlation between years of menopause and peripheral or hepatic insulin sensitivity (4, 5). PCOS is clinically characterized by anovulation and hyperandrogenism; it currently represents one of the most discussed, controversial, and explored areas in reproductive medicine, because this syndrome affects up to 10% of women of reproductive age (6). Although in past years the main requests to gynecologists by PCOS patients concerned an improvement of cause-related hyperandrogenism or a desire for pregnancy, it recently became evident that the metabolic alterations of the syndrome may have important implications for long-term health (6, 7). Because most women with

2 Ciampelli et al. Insulin Resistance in PCOS and Menopause J Clin Endocrinol Metab, March 2005, 90(3): PCOS come to clinical attention when their glucose tolerance is still normal and the cardiovascular damage is presumably at an early stage, screening for insulin resistance would be crucial to identify those subjects at greater risk, allowing appropriate medical intervention; these assumptions are also effective in menopausal population. The gold standard methods to assess insulin sensitivity (euglycemic hyperinsulinemic clamps and minimal model analysis) are time-consuming and difficult to apply in largescale clinical or epidemiological studies, where easier methods are required. This has raised interest in obtaining estimates from glucose and insulin measured in the fasting state or during an oral glucose tolerance test (OGTT). Several indexes have been described, and most of them have been validated with reference methods (8). The aim of the present study was to verify the validity of different indexes of insulin sensitivity in PCOS and menopausal patients by comparing data obtained by these indexes with data from euglycemic-hyperinsulinemic clamp studies. Furthermore, we tried to elaborate cut-off values of insulin resistance for the different indexes, then we analyzed their predictive performance. Subjects and Methods We studied 100 consecutive Caucasian (native Italians) patients with PCOS (age range, yr) attending our divisional outpatient services. All of the women had spontaneous onset of puberty and normal sexual development, and all had oligomenorrhea with chronic anovulation since puberty. All of the women were euthyroid, and none had taken any medication known to affect plasma sex steroids for at least 3 months before the study. PCOS was diagnosed on the basis of clinical findings (the presence of amenorrhea or oligomenorrhea and hirsutism), plasma androgen levels at the upper limit of or above the normal range [androstenedione, ng/ml ( nmol/liter); testosterone, ng/ml ( nmol/liter)], and the presence of bilaterally normal or enlarged ovaries containing at least 7 10 microcysts ( 5 mm in diameter) on ultrasonography. All of the subjects showed an ovarian stroma/ total area ratio greater than 0.34 (9). In approximately 40% of the cases, the diagnosis was also confirmed by laparoscopy. A normal LH/FSH ratio was not considered an exclusion criterion (9). The presence of a late-onset adrenal enzyme defect was excluded by an ACTH test (250 g, iv; Synacthen, Ciba-Geigy, Basel, Switzerland), according to the criteria reported by New et al. (10). We also recruited 110 healthy Caucasian menopausal women aged yr, who were referred to our department for the relief of menopausal symptoms. Women were between 1 and 5 yr postmenopausal; none had undergone hysterectomy or bilateral oophorectomy. Before a woman s enrollment in the study, assessment of plasma FSH [ 50 miu/ml (50 IU/liter)] and 17 -estradiol [ 10 pg/ml ( pmol/liter)] concentrations, mammography, transvaginal ultrasound examination, cervical cytology, and hysteroscopic endometrial biopsy were performed and found to be normal or compatible with menopausal status. No patient was currently taking drugs known to affect lipid or glucose metabolism, nor had they taken any steroid within the previous 6 months. None smoked more than 10 cigarettes per day or drank more than 300 g alcohol/wk. Breast cancer, altered liver or kidney parameters, history of major thromboembolism, thyroid disease, and uncontrolled or -blocker-treated hypertension (systolic blood pressure 160 mm Hg or diastolic pressure 90 mm Hg) were considered exclusion criteria. Each patient gave written informed consent before entering the study; the study protocol had been previously approved by our institutional review board. Studies were conducted on random days; in no case had recent ovulation occurred in the women with PCOS, as evidenced by retrospective measurement of serum progesterone levels on the days of the study. Obesity was defined as a body mass index (BMI) greater than 27 kg/m 2 (11 13) (normal range, kg/m 2 ); a BMI greater than 25 kg/m 2 defined overweight women. Waist circumference was obtained as the minimum value between the iliac crest and the lateral costal margin, whereas hip circumference was determined as the maximum value over the buttocks. The patients were hospitalized before and after a standard carbohydrate diet (300 g/d) for 3 d. After overnight fasting for h, blood samples were collected for basal hormone and lipoprotein assays. Then patients underwent an OGTT. Two days later, after another overnight fast, a euglycemic-hyperinsulinemic clamp was performed. Plasma levels of testosterone, dehydroepiandrosterone sulfate, androstenedione, 17-hydroxyprogesterone, FSH, LH, SHBG, triglycerides, high-density lipoprotein cholesterol, very low density lipoprotein cholesterol, low-density lipoprotein cholesterol, cholesterol, and free fatty acids were determined in basal conditions. The OGTTs (75 g) and euglycemic-hyperinsulinemic clamps were performed as previously described (14, 15). Insulin sensitivity was evaluated as total body glucose utilization (M) expressed as milligrams per kilogram body weight per minute. We preferred this index as the measure of insulin sensitivity, because the M/I ratio fails to narrow the range of individual sensitivity values (15). Samples for hormone assay were promptly centrifuged, and plasma samples were stored at 20 C until assay, whereas samples for biochemical assay were immediately processed. All hormones and lipoproteins were assayed as previously described (9, 14, 16). A normal glycemic response to OGTT was defined according to criteria of the American Diabetes Association (17). The areas under the curve (AUC) at 240 min were evaluated according to the formula: [([V30 V60 V90 V120 (mean V120-V180) V180 (mean V180-V240)] 2) V0 V240] 15, where V is the glucose, insulin, or C-peptide concentration at the indicated time. The AUC at 120 min was evaluated according to the formula: ([(V30 V60 V90) 2] V0 V120) 15. The insulin assay cross-reactivity with proinsulin at the middle curve is approximately 40%. To estimate insulin sensitivity, several indexes from either fasting- or OGTT-derived measurements were used, as shown in Table 1. The results obtained by replacing insulin with C-peptide values were also evaluated for each single index. For the stimulated indexes, formulas were calculated both for 120- and 240-min curves. The presence of insulin resistance was defined as an M value lower than 4.45 mg/kg min calculated by the mean 2 sd of the M values obtained by euglycemic hyperinsulinemic clamps performed in a group of 40 control normoovulatory lean women evaluated in the early follicular phase of their cycle. Normoovulatory cycles were defined according to the presence of progesterone values greater than 7.86 ng/ml (25 nmol/liter) on d 21 of the cycle in three different cycles. Levels of FSH, LH, estradiol, and testosterone in these subjects were all within the normal range of our laboratory. All results are expressed as the mean and range. All variables for evaluation of insulin sensitivity were examined for normality of distribution with the Kolmogorov-Smirnov goodness of fit test. Because of the nonparametric distribution of the analyzed variables, Spearman s rank correlation coefficient (R s ) was used to study the strength of association between measurements of insulin sensitivity. The performance of each index for insulin sensitivity evaluation was described and compared as follows. Receiver operating characteristic (ROC) curves were built and analyzed by using the SPSS release packet (SPSS, Inc., Chicago, IL). A statistical comparison of the ROC AUCs was made according to methods previously described by Hanley and McNeil (29). For each studied group, comparisons were made between the best predictive indexes (the insulin AUC 240 min for the menopausal subjects; the Avignon insulin sensitivity index calculated in basal condition, replacing insulin value with C-peptide value (Table 1 and Ref. 26) for PCOS) and all other parameters considered. Under the hypothesis of normal distribution of the statistical quantity z (29), a one-sided z-test was conducted. Starting from arbitrarily set up cut-off values, ROCs were computed assuming a nonparametric distribution. se, asymptotic significance, and asymptotic 95% confidence intervals were also reported. The positive predictive value (PPV) and the negative predictive value (NPV) were calculated as well. P 0.05 was considered statistically significant.

3 1400 J Clin Endocrinol Metab, March 2005, 90(3): Ciampelli et al. Insulin Resistance in PCOS and Menopause TABLE 1. Indexes of insulin sensitivity derived from fasting and OGTT Index Formula Ref. no. HOMA G mmol/liter I IU/ml 18, Belfiore area 2 20 AUC I IU/ml AUC G mg/dl 1 Belfiore basal a I IU/ml G mg/dl 1 Matsuda and De Fronzo 10, G I Gmeanogtt Imeanogtt Bennet 1 22 ln G mg/dl ln I IU/ml McCauley Esp( ln(i ( IU/ml)) 0.31 ln(trig (mg/dl)) 23 OGIS b 4 pp5 pp6 G 120 Gcl 2 Quicki 1 log I IU/ml log G mg/dl LnHOMA 1 Ln HOMA Sib 100,000,000 I IU/ml G mg/dl 150 ml weight kg Si ,000,000 I120 IU/ml G120 mg/dl 150 ml weight kg SiM Sib Si120 2 Caumo F D AUC G t /G t GE AUC G t /G t AUC G t AUC I t G/I area AUC G mmol/liter 1,000 AUC I pmol/liter G/I basal G mmol/liter 1,000 I pmol/liter G/I secretory area AUC G mmol/liter G mmol/liter t AUC I pmol/liter I pmol/liter t ) I secretory area (AUC(I (pmol/liter)) (I (pmol/liter) t) All OGTT derived indexes were evaluated using both 120- and 240-min curves. All indexes were evaluated also substituting the C-peptide value for the insulin one (these indexes are marked as B in the following tables). Where it is not specified, common or Systeme International units can be used. G, Glucose; I, insulin; C, C-peptide; Trig, triglycerides; t, time; Gmeanogtt, mean glucose value during OGTT; Imeanogtt, mean insulin value during OGTT; G120, glucose value during OGTT at 120 min; I120, insulin value during OGTT at 120 min; F, constant 0.8 (27, 28); D, glucose administrated dose/kg body weight; GE, constant (27, 28);, intermediate calculation (eq. 8 of Ref. 24); pp5 and pp6, intermediate calculation (for more details see Ref. 24). Conversion factors to Systeme International units: glucose, ; insulin, 7.175; C-peptide, 331; triglycerides, a Basal Belfiore s formulas have been multiplied by 1000 to obtain whole numbers. b The formula is also available at // Because determination of the indexes did not significantly differ in most cases, as resulted from the ROC analysis curve, and because different indexes misclassified different patients, we tried to build a linear combination of the considered predictors to obtain a suitable criterion presumably able to predict insulin resistance. Because the hypothesis of normality for the analyzed variables was not confirmed, the discriminant analysis could not be performed, and a logistic regression was applied. Because the predictors strongly correlated, a forward stepwise procedure, based on the likelihood ratio, was used in selecting variables. Following the standard classification, patients were classified as insulin resistant when logit was greater than zero, resulting in a probability greater than 0.5 of being an insulin-resistant case. The two tests were then combined, in series and in parallel, and the new sensitivity and specificity were computed as well as the test efficacy. Results According to American Diabetes Association criteria (17), seven PCOS patients and 13 menopausal subjects had IGT or frank type 2 diabetes mellitus and were excluded from the study. The studied sample thus included 93 PCOS and 97 menopausal women. Table 2 shows clinical, endocrine, and biochemical features of the studied populations; both groups included a wide range of BMI. Fifty-eight of 93 (62.4%) PCOS patients were obese; this was also true for 38 of 97 (39.1%) menopausal women. Table 3 shows several correlation coefficients (R s ) between M values obtained by euglycemic-hyperinsulinemic clamp , 28

4 Ciampelli et al. Insulin Resistance in PCOS and Menopause J Clin Endocrinol Metab, March 2005, 90(3): TABLE 2. Endocrine, clinical, and biochemical parameters of the studied populations PCOS Menopause No. of patients Age (yr) 25 (18 35) 56 (45 60) BMI (kg/m 2 ) ( ) ( ) Waist/hip ratio 0.83 ( ) 0.79 ( ) Fasting glucose (63 109) (70 124) (mg/dl) Fasting insulin ( ) ( ) ( IU/ml) Fasting C-peptide 2.15 ( ) 2.01 ( ) (ng/ml) Testosterone (ng/ml) 0.60 ( ) 0.31 ( ) Androstenedione 2.48 ( ) 1.45 ( ) (ng/ml) DHEAS ( g/ml) ( ) ( ) 17-OHP (ng/ml) 1.49 ( ) 0.89 ( ) SHBG (nmol/liter) ( ) ( ) Estradiol (pg/ml) (13 345) 10 LH (miu/ml) 8.63 ( ) ( ) FSH (miu/ml) 5.40 ( ) ( ) Cholesterol (mg/dl) ( ) ( ) Triglycerides (mg/dl) (28 265) (40 236) LDL (mg/dl) (51 173) (71 205) HDL (mg/dl) (21 75) (27 81) VLDL (mg/dl) (6 53) (8 127) FFA (meq/liter) 0.44 ( ) 0.38 ( ) Data are presented as the mean (range). FFA, Free fatty acids; DHEAS, dehydroepiandrosterone sulfate; HDL, high-density lipoprotein; LDL, low-density lipoprotein; 17OHP, 17-hydroxyprogesterone; VLDL, very low density lipoprotein. Conversion factors for Systeme International units: glucose, ; insulin, 7.175; C-peptide, 331; testosterone, 3.467; androstenedione, 3.49; DHEAS, 2.714; 17OHP, 3.180; estradiol, 3.671; triglycerides, ; cholesterol; LDL, HDL, vldl, studies and the indexes obtained by fasting or stimulated insulin, C-peptide, and glucose values. The R s values are presented in decreasing order within PCOS patients; the corresponding correlation coefficients for menopausal subjects are also reported in Table 3 (right side). Only the most significant data are reported in the table, although almost all of the evaluated indexes showed a statistically significant correlation with the M value. In particular, the original formula of homeostasis model assessment (HOMA) showed an R s of in PCOS and of in menopausal subjects. The R s values for fasting insulin were and , respectively. The glucose to insulin ratio of the AUC (G/I area) and in the fasting state (G/I basal) showed R s of and , respectively, in PCOS women and values of and in menopausal subjects. The data for BMI showed correlation coefficients of and , respectively, in PCOS and menopausal women. Correlation coefficients were generally higher in PCOS than menopausal women. Furthermore, there was not complete agreement between the best correlated indexes in PCOS and menopausal subjects; indeed, the best R s values within PCOS group were found for Avignon insulin sensitivity SiM index (26) and Avignon insulin sensitivity basal index (Sib) (26). In a similar evaluation within a menopausal population, the highest R s values were instead found for the formulas reported by Matsuda and De Fronzo (21) and Belfiore et al. (20), calculated by 120- or 240-min curves. We found superimposable results in evaluating the formulas at 120- or 240- min curves. For example, the insulin AUC R s in PCOS was at 240 min and at 120 min. For menopausal patients, the R s between the M value and the insulin AUC was at 240 min and at 120 min. Thus, only the formulas calculated by 120-min curves are reported in Table 3. According to the cut-off of 4.45 mg/kg min (M value) for the diagnosis of insulin resistance evaluated by euglycemichyperinsulinemic clamp studies, 66.1% of PCOS women and 44.3% of menopausal women were classified as insulin resistant. As described above, using arbitrary initial cut-off values for insulin resistance, we examined the sensitivity and specificity of various cut-off values for these indexes. Figure 1 shows the ROC graphs, where sensitivity is plotted against 1 specificity (or the false positive rate). The ideal screening test is supposed to approach or reach the upper left corner of the graph (100% sensitivity and 100% specificity). A test that approximates a coin flip is the diagonal from the lower left to the upper right corner of the graph. The cut-off point for each of the screening tests that has the best combination of sensitivity and specificity is located at the knee of the graph. The predictive performance of the best cut-off values obtained for the investigated indexes of insulin resistance are shown in Tables 4 and 5; sensitivity, specificity, PPV, and TABLE 3. Spearman s correlation coefficients (R s ) between several indexes of insulin sensitivity and the insulin-mediated glucose disposal obtained by euglycemic hyperinsulinemic clamp test Index PCOS Menopause SiM R s P R s P Sib B a R s P R s P SiM B a R s P R s P Sib R s P R s P Si120 b R s P R s P Insulin area ( IU/ml 240 min) R s P R s P Matsuda, De Fronzo 120 min b R s P R s P G/I basal R s P R s P Belfiore basal B a R s P R s P HOMA B a R s P R s P Insulin area ( IU/ml 120 min) R s P R s P Belfiore area 120 min b R s P R s P For more details about the abbreviations, see Table 1. a The formulas marked as B are calculated using C-peptide instead of insulin values. b Formulas calculated by 120-min curves.

5 1402 J Clin Endocrinol Metab, March 2005, 90(3): Ciampelli et al. Insulin Resistance in PCOS and Menopause FIG. 1. ROC analysis of different fasting- or OGTT-derived parameters of insulin resistance in PCOS (A and B) and menopausal patients (C and D). Belfiore area, Stimulated Belfiore index; Belfiore basal B, fasting Belfiore index calculated by replacing C-peptide with insulin values in the original formula; HOMA B, HOMA calculated by replacing C-peptide with insulin values in the original formula; G/I basal, ratio G/I calculated at the fasting blood value of glycemia and insulinemia; I area 240, insulin AUC calculated at 240 min; I area 120, insulin AUC calculated at 120 min; Matsuda De Fronzo 120, formula calculated by 120-min curve; Sib, insulin sensitivity calculated by fasting values; Si120, insulin sensitivity calculated at 120 min after the glucose load; SiM, insulin sensitivity calculated as intermediate value between Sib and Si120; Sib B and SiM B, Sib and SiM calculated by replacing C-peptide with insulin values. For more details about the formulas used, see Table 1. NPV are reported. Furthermore, all statistical analysis concerning the ROC curves for each index have been reported; tables specify the area value, sd, confidence interval, and P value between several indexes. The best predictive performance in diagnosing insulin resistance was obtained in PCOS subjects using a SibB (Sib calculated by glucose and C-peptide values) value of 62 or less (AUC ROC 87.66%), an insulin AUC of 7,000 IU/ml or more ( 50,225 pmol/ liter) 120 min (AUC ROC 87.26%), and an Si120 of 2 or less (AUC ROC 84.57%). In menopausal subjects the best predictive indexes of insulin resistance were an insulin AUC of 10,000 IU/ml or more ( 71,750 pmol/liter) 240 min (AUC ROC 79.37%) and a Belfiore area calculated at 120 min of 0.85 or less (AUC ROC 77.01%). For the menopausal subjects, the results of all tests, detecting possible differences between the ROC AUC of the best predictive index (insulin AUC at 240 min) and the areas related to the other parameters were nonsignificantly different, except for the G/I ratio, which showed a lower predictivity (P 0.04). For the PCO group, significant differences

6 Ciampelli et al. Insulin Resistance in PCOS and Menopause J Clin Endocrinol Metab, March 2005, 90(3): TABLE 4. Predictivity of parameters of insulin resistance in PCOS patients Index arose between both SibB and Sib (P 0.03), G/I ratio (P 0.02), and Belfiore basal (P 0.001; Table 4). In PCOS women, the only significant factors were SiM and insulin AUC at 240 min. Results of the logistic regression are reported in Table 6. According to this classification, seven cases were misclassified (two insulin-resistant patients were coded as insulin sensitive, and five insulin-sensitive patients were coded as resistant, giving rise to a specificity of 73.68% and a sensitivity of 94.59%) compared with nine misclassified cases (eight insulin-resistant patients were coded as insulin sensitive, and one insulin-sensitive patient was coded as resistant) when the performance of the SibB, the best test in the ROC analysis, was taken into account. The two tests are not independent, because they are based on strongly correlated measures, so the use of the tests in combination, in series or in parallel, does not provide very different results. The greatest test efficacy was obtained for the combined parallel test (89.29%). Table 7 provides detailed results. For menopausal patients, the variables included in the final TABLE 5. Predictivity of parameters of insulin resistance in menopause patients Index Cut-off Cut-off Sens Sens Spec Spec PPV PPV model were SibB and insulin-auc at 240 min (for details, see Table 8). According to the logistic regression test, 14 cases were misclassified (nine insulin-resistant patients were coded as insulin sensitive, and five insulin-sensitive patients were coded as resistant, giving rise to a specificity of 87.5% and a sensitivity of 71.87%), whereas 15 cases were misclassified when the classification was made according to insulin AUC at 240 min, the variable that produced the highest ROC AUC (10 insulin-resistant patients were coded as insulin sensitive, and five insulinsensitive patients were coded as resistant). The highest efficacy was obtained with the tests combined in series (81.94%; Table 9); however, very slight differences among the tests were found. Our results lead to a preference for the combined tests, because the increase in performance is so slight that it does not warrant the more cumbersome procedure. Discussion The maintenance of normal glucose homeostasis involves the simultaneous and coordinated roles of pancreatic -cells, NPV Area SE Asympt signif 95% Confidence interval Belfiore area 120 min a Belfiore basal B b HOMA B b G/I basal Insulin area ( IU/ml 240 min) 10, c Insulin area ( IU/ml 120 min) 7, Matsuda, De Fronzo 120 min a Sib Si SiM Sib B b SiM B b Sens, Sensitivity; Spec, specificity. For more details about the abbreviations, see Table 1. a Formulas calculated by 120 min curves. b The formulas marked B are calculated using C-peptide instead of insulin values. c P values were calculated with respect to insulin area 240 min. NPV Area SE Asympt signif 95% Confidence interval Belfiore area 120 min a Belfiore basal B b HOMA B b G/I basal Insulin area ( IU/ml 240 min) 15, Insulin area ( IU/ml 120 min) 7, Matsuda, De Fronzo 120 min a Sib Si120 a SiM Sib B b c SiM B b Sens, Sensitivity; Spec, specificity. For more details about the abbreviations, see Table 1. a Formulas calculated by 120-min curves. b The formulas marked B are calculated using C-peptide instead of insulin values. c P values were calculated with respect to Sib B. Lower bound Lower bound Upper bound Upper bound P value P value

7 1404 J Clin Endocrinol Metab, March 2005, 90(3): Ciampelli et al. Insulin Resistance in PCOS and Menopause TABLE 6. Results of the logistic regression for the PCOS population Parameters in the model SE P SiM Insulin area 240 min Constant TABLE 8. Results of the logistic regression for the menopausal population Parameters in the model SE P Insulin area 240 min E SibB Constant liver, and peripheral tissues, primarily muscle (30). During the last 20 yr, the assessment of in vivo insulin sensitivity in humans has been frequently based on the use of glucose clamp technique (14). Although it does not mirror the physiological condition of continuously changing glucose and insulin levels and the different insulin exposure in the liver and peripheral tissues, the glucose clamp technique is the method with the fewest drawbacks, yielding results closest to the real measure; this technique is thus considered the gold standard and the reference method to evaluate insulin resistance (21, 31). However, the glucose clamp is not easily applied in large scale investigations, because iv infusion of insulin, frequent blood samples, and continuous adjustment of a glucose infusion are required for each subject studied. A well-accepted alternative to the clamp technique was the iv glucose tolerance test interpreted with the classic minimal model of glucose disappearance (32, 33). This method has gained increasing popularity over the years (34); however, although this approach is less labor intensive than the glucose clamp, the minimal model is still not ideal for large studies because approximately 30 blood samples over 3 h are required. Furthermore, there is not general agreement about the levels of correlation obtained between direct measures of insulin sensitivity (i.e. glucose clamp) and indirect measures, such as minimal model analysis, with correlation coefficients varying from (25, 35, 36). These data suggest that investigators should be cautious in applying minimal model analysis of insulin sensitivity to population studies. This is highlighted by recent studies demonstrating particular inadequacies of the minimal model approach that result in overestimation of glucose effectiveness and underestimation of insulin sensitivity (34, 37). Alternative methods applicable to large studies have been proposed for measurement of insulin sensitivity (14 28), such as fasting insulin and glucose as well as stimulated (OGTT) plasma levels. Indeed, although it is well known that OGTT provides information about insulin secretion and action, but does not directly yield a measure of insulin sensitivity, several formulas have been proposed and successfully tested against the clamp (20, 21, 24, 26 28). The significance of a valid index of insulin sensitivity as a screening method in the primary prevention of diabetes or CVD in the gynecological populations analyzed in the present study is suggested by two observations. First, these TABLE 7. Test results for the PCOS population Test Sensibility Specificity Efficacy ROC analysis test Logistic regression test Combined test in series Combined test in parallel patients are characterized by increased type 2 diabetes mellitus or CVD risk (6, 7, 38, 39). Furthermore, our data indicate a higher prevalence of insulin resistance in PCOS and menopausal women (66.1% and 44.3%, respectively) compared with the 20 25% insulin-resistant subjects found in the nondiabetic general population (21) (a limitation could be identified in the selection of menopausal women, because most of them were referred to our clinical service for the relief of menopausal symptoms). In the present work we analyzed almost all the known formulas for insulin sensitivity measurement; for each index we also evaluated the results obtained by replacing C-peptide to insulin values, because C-peptide assays are less influenced by possible assay interactions, such as that between insulin and proinsulin. The subjects already diagnosed as IGT or type 2 diabetes mellitus patients were excluded from our analysis because we wanted to validate a screening method for primary prevention. Our results showed that the indexes of insulin sensitivity best correlated to the clamp studies were, for the PCOS population, those processed by Avignon et al. (26); in particular, SiM reached an R s of , whereas evaluation in the fasting state using C-peptide instead of insulin concentrations reached an R s of In contrast, in the menopausal population the best correlation coefficients were obtained for the formulas reported by Belfiore et al. (20) and Matsuda and De Fronzo (21) (R s and R s for Belfiore index evaluated by 240- or 120-min curves; R s and R s for a similar evaluation by the formula proposed by Matsuda and De Fronzo). We generally found quite superimposable correlation coefficients by using curves at 120 or 240 min; these data suggest the usefulness of the 120-min curves. Concerning the HOMA index, our results showed a weaker association with the clamp-measured total glucose disposal (R s 0.57 in PCOS, R s 0.46 in menopausal women) compared with other correlation coefficients reported previously ( ) (16, 17, 19, 24). However, our data are in line with those reported by McAuley et al. (23) (R s 0.53) in 178 normoglycemic individuals and by Anderson et al. (R s 0.40) (22). These discrepant data cannot be easily explained; it is possible that our results are influenced by the selection of two different populations with a high prevalence of insulin resistance. In any case, our data do not TABLE 9. Test results for the menopausal population Test Sensibility Specificity Efficacy ROC analysis test Logistic regression test Combined test in series Combined test in parallel

8 Ciampelli et al. Insulin Resistance in PCOS and Menopause J Clin Endocrinol Metab, March 2005, 90(3): seem to validate the use of a simple HOMA for screening insulin resistance in PCOS and menopausal subjects. Although many indexes have been tested for the evaluation of insulin sensitivity, with different correlation coefficients compared with clamp or minimal model analysis (18 28), a flaw in the reported studies concerns the lack of a predictive performance of cut-off values for the analyzed formulas. The introduction of indexes allowing an easy determination of insulin resistance is aimed at replacing the euglycemic-hyperinsulinemic clamp and the minimal model analysis, so there is a need for cut-off values characterized by a good predictive performance to identify insulin-resistant subjects. Only McAuley et al. (23), in analyzing a weighted combination of fasting insulin and triglycerides, reported a sensitivity of 62% and a specificity of 84% in identifying insulin resistance. Legro et al. (40) tried to validate a screening test for insulin resistance in women with PCOS; the researchers found that setting a value of the fasting G/I ratio of less than 4.5 as abnormal, the sensitivity of this parameter was 95%, the specificity was 84%, the PPV was 87%, and the NPV was 94%; these data suggest very interesting results in terms of predictive performance. However, some personal comments arose from the data presented. Only obese (n 40) PCOS patients were studied; thus, the results cannot be validated in a general PCOS population. The researchers themselves commented that the fasting G/I ratio would not be predicted to be a reliable measure of insulin resistance in nonobese PCOS women because they have neither fasting hyperinsulinemia nor increased basal hepatic glucose production. Another important flaw found in the report by Legro consists of the definition of insulin resistance; in fact, the researchers defined abnormal insulin sensitivity values as the 10th percentile for insulin sensitivity index found in 15 obese controls, because they wanted to assess insulin resistance that was beyond that due to obesity per se. In our opinion this could not be defined as insulin resistance, as also suggested by the presence of normal insulin sensitivity in 47% of obese PCOS women. In our series, the best predictive indexes of insulin resistance in PCOS were the insulin AUC and those proposed by Avignon et al. (26). Within menopausal subjects, the best predictivity was obtained by the insulin AUC calculated by 240-min curves and the Belfiore area (20) index calculated by 120-min curves. Several important considerations might be drawn from our results. Firstly, even if there are higher correlation coefficients for some indexes, there is no really significant difference in terms of predictive performance between the simple insulin AUC and the other reported indexes. Furthermore, the cut-off values for each variable are not completely superimposable between PCOS and menopausal women, suggesting the need to establish different cut-off values for different populations. We urge caution in the application of the insulin AUC or other indexes in monitoring the incidental effects of diet or insulin-lowering drug administration on insulin resistance. Indeed, we previously showed (41) that in PCOS subjects the administration of an oral opioid antagonist (naltrexone) for about 1 month was able to reduce the exaggerated insulin secretion after a glucose load by about 30% without affecting the glucose level. An improvement of insulin sensitivity in these patients could be expected; by contrast, data from euglycemic-hyperinsulinemic clamp studies did not show any modification in the insulin-mediated glucose disposal, yielding the conclusion that naltrexone might act at the hepatic level (insulin extraction). In conclusion, our data show that several indirect indexes of insulin resistance obtained from fasting or stimulated glucose, insulin, or C-peptide concentrations are superimposable, in terms of strength of association with the clamp data, to those reported by other researchers for the minimal model (25, 35, 36), suggesting that they can effectively replace the minimal model or clamp evaluations in large population studies, in particular those involving PCOS or menopausal women. The presence of a high correlation coefficient does not mean that these indexes have the best predictive performance in diagnosing insulin resistance, because of the presence of many borderline values. Our data suggest that different indexes of insulin resistance could be indifferently used to establish insulin resistance, because there is no statistically significant difference in the ROC curve analysis. However, we believe that it is better to use the one with the highest ROC curve area, because it can be more representative of all populations, even considering its PPV. Thus, the analysis of predictivity showed that the simple evaluation of insulin AUC seems to be the best predictive index of insulin resistance in PCOS and menopausal patients. Acknowledgments Received March 1, Accepted December 2, Address all correspondence and requests for reprints to: Dr. Antonio Lanzone, Department of Obstetrics and Gynecology, Università Cattolica del Sacro Cuore, L.go A. Gemelli 8, Rome, Italy. alanzone@rm.unicatt.it. References 1. Moller DE, Flier JS 1991 Insulin resistance-mechanisms, syndromes, and implications. N Engl J Med 325: Reaven GM 1988 Role of insulin resistance in human disease. Diabetes 37: Laws A, Reaven GM 1993 Insulin resistance and risk factors for coronary heart disease. In: Ferrannini E, eds. Clinical endocrinology and metabolism. London: Bailliere Tindall; Lindheim SR, Presser SC, Ditkoff EC, Vijod MA, Stanczyk FZ, Lobo RA 1993 A possible bimodal effect of estrogen on insulin sensitivity in postmenopausal women and the attenuating effect of added progestin. Fertil Steril 60: Godsland IF, Crook D, Stevenson JC, Collins P, Rosano GMC, Less B, Sidhu M, Poole-Wilson PA 1995 The insulin resistance syndrome in postmenopausal women with cardiological syndrome X. Br Heart J 74: Ciampelli M, Lanzone A 1998 Insulin and polycystic ovary syndrome a new look at an old subject. Gynecol Endocrinol 12: Rajkhowa M, Glass MR, Rutheford AJ, Michelmore K, Balen AH 2000 Polycystic ovary syndrome: a risk factor for cardiovascular disease? Br J Obstet Gynecol 107: Radziuk J 2000 Insulin sensitivity and its measurement: structural commonalities among the methods. J Clin Endocrinol Metab 85: Fulghesu AM, Ciampelli M, Belosi C, Apa R, Pavone V, Lanzone A 2001 A new ultrasound criterion for the diagnosis of polycystic ovary syndrome: the ovarian stroma/total area ratio. Fertil Steril 76: New MI, Lorentzen F, Lerner AJ, Kohn B, Oberfield SE, Pollack Ms, Dupont B, Stoner E, Levy DJ, Pang S, Levine LS 1983 Genotyping steroid 21-hydroxylase deficiency: hormonal reference data. J Clin Endocrinol Metab 57: Morin-Papunen L, Rautio K, Ruokonen A, Hedberg P, Puukka M, Tapanainen JS 2003 Metformin reduces serum C-reactive protein levels in women with polycystic ovary syndrome. J Clin Endocrinol Metab 88: Koivunen RM, Morin Papunen LC, Ruokonen A, Tapanoinen JS, Martikoinen HK 2001 Ovarian steroidogenic response to human chorionic gona-

9 1406 J Clin Endocrinol Metab, March 2005, 90(3): Ciampelli et al. Insulin Resistance in PCOS and Menopause dotrophin in obese women with polycystic ovary syndrome: effect of metformin. Hum Reprod 16: Calle EE, Thun MJ, Petrelli JM, Rodriguez C, Heath Jr CW 1999 Body-mass index and mortality in a prospective cohort of U.S. adults. N Engl J Med 341: Ciampelli M, Fulghesu AM, Cucinelli F, Pavone V, Caruso A, Mancuso S, Lanzone A 1997 Heterogeneity in cell activity, hepatic insulin clearance and peripheral insulin sensitivity in women with polycystic ovary syndrome. Hum Reprod 12: Bergman, RN, Finegood, DT Ader, M 1985 Assessment of insulin sensitivity in vivo. Endocr Rev 6: Ciampelli M, Fulghesu AM, Cucinelli F, Pavone V, Ronsisvalle E, Guido M, Caruso A, Lanzone A 1999 Impact of insulin and body mass index on metabolic and endocrine variables in polycystic ovary syndrome. Metabolism 48: Expert Committee on the Diagnosis and Classification of Diabetes Mellitus 1997 Report of the expert committee on the diagnosis and classification of diabetes mellitus. Diabetes Care 20: Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF, Turner RC 1985 Homeostasis model assessment: insulin resistance and -cell function from fasting plasma glucose and insulin concentrations in man. Diabetologia 28: Bonora E, Targher G, Alberiche M, Bonadonna RC, Saggiani F, Zenere MB, Monauni T, Muggeo M 2000 Homeostasis model assessment closely mirrors the glucose clamp technique in the assessment of insulin sensitivity. Diabetes Care 23: Belfiore F, Iannello S, Volpicelli G 1998 Insulin sensitivity indices calculated from basal and OGTT-induced insulin, glucose and FFA levels. Mol Genet Metab 63: Matsuda M, De Fronzo RA 1999 Insulin sensitivity indices obtained from oral glucose tolerance testing. Diabetes Care 22: Anderson RL, Hamman RF, Savage PJ, Saad MF, Laws A, Kades WW, Sand RE, Cefalu W 1995 Exploration of simple insulin sensitivity measures derived from the frequently sampled intravenous glucose tolerance (FSIGT) tests. Am J Epidemiol 142: McAuley KA, Williams SM, Mann JI, Walker RJ, Lewis-Barned RJ, Temple LA, Duncan AW 2001 Diagnosing insulin resistance in the general population. Diabetes Care 24: Mari A, Pacini G, Murphy E, Ludvik B, Nolan JJ 2001 A model-based method for assessing insulin sensitivity from the oral glucose tolerance test. Diabetes Care 24: Katz A, Nambi SS, Mather K, Baron AD, Follmann DA, Sullivan G, Quon MJ 2000 Quantitative insulin sensitivity check index: a simple accurate method for assessing insulin sensitivity in humans. J Clin Endocrinol Metab 85: Avignon A, Boegner C, Mariano-Goulart D, Colette C, Monnier L 1999 Assessment of insulin sensitivity from plasma insulin and glucose in the fasting or post oral glucose-load state. Int J Obes 23: Caumo A, Bergman RN, Cobelli C 2000 Insulin sensitivity from meal tolerance tests in normal subjects: a minimal model index. J Clin Endocrinol Metab 85: Breda E, Cavaghan MK, Toffolo G, Polonsky KS, Cobelli C 2001 Oral glucose tolerance test minimal model indexes of -cell function and insulin sensitivity. Diabetes 50: Hanley JA, McNeil BJ 1983 A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology 148: De Fronzo RA 1988 The triumvirate: -cell, muscle, liver: a collusion responsible for NIDDM. Diabetes 37: Myllynen P, Koivisti V, Nikkila E 1987 Glucose intolerance and insulin resistance accompany immobilization. Acta Med Scand 222: Bergman RN, Ider YZ, Bowden CR, Cobelli C 1979 Quantitative estimation of insulin sensitivity. Am J Physiol 236:E667 E Bergman RN, Phillips NLS, Cobelli C 1981 Physiologic evaluation of factors controlling glucose tolerance in man. Measurement of insulin sensitivity and -cell sensitivity from the response to intravenous glucose. J Clin Invest 68: Caumo A, Vicini P, Cobelli C 1996 Is the minimal model too minimal? Diabetologia 39: Donner CC, Fraze E, Chen YD, Hollenbeck CB, Foley JE, Reaven GM 1985 Presentation of a new method for specific measurement of in vivo insulinstimulated glucose disposal in humans: comparison of this approach with the insulin clamp and minimal model techniques. J Clin Endocrinol Metab 60: Swan JW, Walton C, Godsland IF 1994 Assessment of insulin sensitivity in man: a comparison of minimal model- and euglycemic clamp-derived measures in health and heart failure. Clin Sci 86: Saad MF, Anderson RL, Laws A, Watanabe RM, kades WW, Chen YD, Sands RE, Pei D, Savage PJ, Bergman RN 1994 Insulin Resistance Atherosclerosis Study: a comparison between the minimal model and the glucose clamp in the assessment of insulin sensitivity across the spectrum of glucose tolerance. Diabetes 43: Hjortland MC, McNamara PM, Kannel WB 1976 Some atherogenic concomitants of the menopause. The Framingham study. Am J Epidemiol 103: Matthews K, Meilahn E, Kuller L, Kelsey S, Caggiula A, Wing R 1989 Menopause and risk factors for coronary heart disease. N Engl J Med 321: Legro RS, Finegood D, Dunaif A 1998 A fasting glucose to insulin ratio is a useful measure of insulin sensitivity in women with polycystic ovary syndrome. J Clin Endocrinol Metab 83: Fulghesu AM, Ciampelli M, Guido M, Murgia F, Caruso A, Mancuso S, Lanzone A 1998 Role of the opioid tone in the pathophysiology of hyperinsulinemia and insulin resistance in polycystic ovarian disease. Metabolism 47: JCEM is published monthly by The Endocrine Society ( the foremost professional society serving the endocrine community.